Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Author:
NAZARIS EVANGELOS
Supervisors info:
Γουνόπουλος Δημήτριος, Καθηγητής, ΤΜΗΜΑ ΠΛΗΡΟΦΟΡΙΚΗΣ ΚΑΙ ΤΗΛΕΠΙΚΟΙΝΩΝΙΩΝ, ΕΘΝΙΚΟ ΚΑΙ ΚΑΠΟΔΙΣΤΡΙΑΚΟ ΠΑΝΕΠΙΣΤΗΜΙΟ ΑΘΗΝΩΝ ΣΧΟΛΗ ΘΕΤΙΚΩΝ ΕΠΙΣΤΗΜΩΝ
Original Title:
Object detection with computer vision Algorithm for car detection
Translated title:
Object detection with computer vision Algorithm for car detection
Summary:
Our project is to make a car detector that warns you if you get close to a car.Detectors scan video or images and make a square around the car. Its square color represents the car distance.
Our research depends on testing some basic and new algorithms in car detections. Type of algorithms that we use are region based convolutional neural networks. We test the time and accuracy that algorithms can detect an object. Two algorithms that we use are RCNN and faster RCNN. We take an unexpected result and the faster RCNN has more accuracy than the RCNN algorithm . I say unexpected because faster RCNN makes prediction for finding an object while RCNN algorithm scans all the pictures. So faster RCNN is 146 times faster than RCNN.
We also discover that with gray filters the algorithms run better.
To start with, we will make a small introduction to the machine learning methodology, and the Object recognition with Deep Learning. Consequently, we will make a brief presentation of the algorithms used in car detection mainly focusing on the Faster R-CNN model (the model based on which we constructed our very own car detector).At the end of this paper we will saw you samples of our code execution that explain our project properly
Main subject category:
Technology - Computer science
Keywords:
Object detection, computer vision, Algorithm for car detection, neural networks , R-CNN, Faster R-CNN,Gray filter on car detection,SSD,YOLO.
